We report on a passive all-optical clock recovery technique based on data signal filtering with a Fabry-Perot filter, tested in a 40 Gb/s transmission system. We have simulated the clock recovery principle to choose the filter finesse and then investigate with experiment the method for 43 Gbit/s RZ signal clock recovery ahead of a receiver. We use Bit Error Rate assessment to demonstrate its system compatibility and to evaluate both its pattern sequence length tolerance and, for the first time, its clock locking range.
© 2007 Optical Society of America
All Optical Clock Recovery (OCR) is a key function for optical signal processing: 40 Gbit/s transmissions and above , , optical packet switched networks , optical regeneration  and optical demultiplexing .
A full Bit Error Rate (BER) assessment with a simple clock recovery scheme  based on a commercial high finesse Fabry-Perot Filter (FPF) followed by a Semiconductor Optical Amplifier (SOA) power equalizer  has been recently reported at 43 Gbit/s . All Optical Clock Recovery passive techniques appear very attractive for its simplicity and its compatibility with all optical signal processing at 40 Gbit/s and beyond . Moreover several active techniques has been recently demonstrated at high bit rate, using externally or internally mode-locked devices, with better timing jitter performances but more complicated implementation , .
In this article, we present a full investigation of this clock recovery scheme in a data transmission system environment, ahead of a receiver, to demonstrate its system compatibility. To this end, the performance with optical clock has been compared to the reference clock from the transmitter and with a fully optoelectronic OCR . Some sensitivity to pattern sequence length of the OCR has been observed. The tolerance of the receiver to the recovered clock quality and its locking range have been also evaluated
2. Simulation of clock recovery set-up
The well known clock recovery scheme used is shown in Fig. 1. It consists of a FPF whose Free Spectral Range (FSR) is equal to the residual line frequency separation of the 43 GHz RZ input data spectrum. The clock recovery technique is a narrowband filtering of the signal modulation lines. The SOA is then used as a power equalizer for reducing the residual amplitude clock modulations due to light polarization effects or mismatching between the FSR and the line spacing in the optical spectrum.
In order to specify the characteristics of the optical clock recovery, we firstly carried out a numerical simulation of the filtering process, using Optisystem commercial software. For this simulation we used a pattern sequence length of 29 (512 bits per sequence) and 8 ps RZ data signal at 40 Gbit/s with pulse shape characteristics close from the experimental one. Varying the filter finesse allowed us to observe a characteristic degradation of the clock quality due to long sequences of spaces . This degradation is evaluated using an original technique based on the re-modulation of the recovered clock through a coder driven by the electrical data from the transmitter. Eye Opening factor is used to compare the quality of the re-modulated clock to the input data signal from the transmitter. By this way it is possible to evaluate the degradation introduced through the filter and also the enhancement brought through the SOA. This technique has been recently proposed for experimental characterization of optically recovered clocks .
The simulation has been performed with the parameters given in the Table 1. Most of those used for the simulation correspond to the Optisystem default parameters. The SOA’s model used in this software is the classical Agrawal model .
Estimation of the Eye Opening factor as a common measurement for optical transmission quality is able to show the degradation of the signal when varying the FPF finesse value. We measured it on the re-modulated clock, firstly after the FPF, and then after the SOA, to compare them to the transmitted signal. The Eye Opening factor value of the input signal (without degradation) is 1 and corresponds to an extinction ratio of more than 15 dB.
It is then possible to plot the curve shown in Fig. 2, the Eye Opening factor versus the Fabry-Perot filter finesse. The insets in Fig. 2 present the eye diagram of the re-modulated clock for each configuration, for a FPF finesse value of 500. With this simulation result we show clearly that FFP output clock quality is enhanced through the SOA by reducing amplitude fluctuations. We show that the Eye Opening factor estimated on the re-modulated recovered clock is also enhanced when increasing the FPF finesse and tends towards 1.
In order to get, in the following experimental part, an optical clock quality suitable for optical system application, we chose a Fabry-Perot finesse corresponding in the simulation, to an Eye Opening factor value close to 1. This finesse value is about 500 and corresponds to a Fabry-Perot linewidth of about 100 MHz at 40 GHz. This spectral range corresponds to natural wavelength fluctuations of a standard DFB laser used in optical telecommunications. This point allows to consider the technique completely compatible with an experimental validation.
3. Experimental set-up
For the experimental set-up, we used a commercial Fabry-Perot filter (FFP-I from Micron Optics) with 43.2 GHz of FSR. The finesse value requested to the manufacturer was 500. The delivered filter finesse value has been experimentally measured to 546 +/- 20. The SOA is a commercial module from Alcatel which operated at an input power level of - 10 dBm. When used like this, the SOA operates as a high pass optical filter, which is able to reduce low frequency optical fluctuations.
In Fig. 3, we show the experimental set-up used for the optical clock recovery evaluation in a system environment.
The transmitter block consists of an Anritsu MP1758 pattern gene rator which is synchronised by an external variable frequency synthesizer at N GHz, a 4:1 electrical multiplexer SHF 404 to obtain a 4×N Gbit/s electrical signal and a ×2 multiplier to generate a 2×N GHz electrical clock. The optical signal from the laser is firstly modulated through a Lithium Niobate modulator (Shaper) driven, with the electrical reference clock to obtain an optical clock with 8 ps FWHM pulses at 4×N GHz. Next, the data are coded on this optical clock through a second modulator (carver) to obtain the data optical signal at 4×N Gbit/s (shown in the left hand inset eye diagram in Fig. 3) with a typical extinction ratio of 15 dB. A very stable laser (OFS 300 from Dicos) with a wavelength fluctuation of a few Megahertz was used in all the experiment.
The receiver block consists of a SHF 423 electrical demultiplexer driven by a 4×N Gbit/s signal and a 2×N GHz electrical clock, followed by an Anritsu MP1776 error detector at N Gbit/s.
The Clock Recovery section consist of three elements: the optical clock recovery scheme we described earlier, a 50 GHz bandwidth photodetector which converts the optical clock to an electrical clock at 4×N GHz and finally, a divide by two frequency divider to reduce the electrical clock from 4×N GHz to the 2×N GHz required by the electrical demultiplexer.
4. Results and set-up comparison
To assess the clock recovery performance, we measured the Bit Error Rate (BER) versus the Optical Signal to Noise Ratio (OSNR) in a bandwidth of 1 nm. The OSNR could be degraded by the classical technique of the addition of ASE noise to the data.We tested the receiver using the degraded signal and an electrical clock. The electrical clocks used were 1) the direct electrical reference clock and 2) the optically recovered clock followed by a digital by two frequency divider as shown in Fig. 3.
Figure 4 shows clearly that the receiver performance does not depend on the clock used: performance is almost identical whether the reference or optically recovered clock is used. In both case a sequence length of 231–1 was employed. Similar results were obtained using an analogue frequency divider.
We studied BER penalty evolution versus input signal frequency detuning in order to investigate the performance of the optical clock recovery ahead of an electrical receiver. For this experiment, we compared two different frequency dividers to reach a frequency suitable for driving the RX block. One is an analogue phase lock loop and the other a commercial digital frequency divider. Figure 5 shows the relative receiver penalty when the data frequency is detuned from the 43.2 GHz nominal FPF FSR. For the analogue frequency divider, we measured a detuning range larger than 200 MHz. For the digital frequency divider, we measured a detuning range of about 40 MHz. These results show a considerable difference depending on electrical set-up ahead of the receiver. This performance difference is due to the filtering effect on the analogue divider which has the effect of increasing the overall quality factor of the clock recovery subsection. In addition, logic components as the digital divider appear to be more sensitive to the quality of the recovered optical clock and, in particular, to its contrast. Cut-off frequency and timing jitter analysis of this OCR device has been studied in system environment with an original technique based on data re-modulation of the optically recovered clock .
Finally, we investigated BER dependence on pattern sequence length which is known to have an influence on optically recovered clock . Results in Fig. 6 show clearly a penalty of 2 dB between 27–1 and 215–1 sequences with almost the same penalty at 215–1. For this experiment, we use the set-up described in Fig. 3 with an analogue frequency divider and the data modulation frequency slightly shifted from the optimal operating point.
We show for the fist time with BER assessment at 40 Gbit/s stability and robustness of this type of optical clock recovery in a data transmission system experiment.
We report a complete study of a simple All Optical Clock Recovery used directly ahead of a receiver compared to an ideal electrical reference.
These results show that it is possible to use an optically recovered clock at 43 Gb/s to synchronise a transmission receiver.
On the other hand, the quality of the electrical clock obtained from the optically recovered clock, depends on electronic processing such as frequency division or data demultiplexing. Therefore, this is not the best way to characterize the optical clock recovery function.
The authors are especially grateful to all the partners of the ROTOR project for sharing their expertise and fruitful discussions. System results have been achieved in the PERSYST Platform of CNRS FOTON-ENSSAT laboratory. The authors would like to thank the Region Bretagne and the Ministère de la Recherche for their support.
References and links
1. A. D. Ellis, K. Smith, and D. M. Patrick, “All Optical Clock Recovery at bit rates up to 40 Gbit/s,” Electron. Lett. 29, 1323 (1993). [CrossRef]
2. D. T. K. Tong, K.-L. Deng, B. Mikkelsen, G. Raybon, K. F Dreyer, and J. E. Johnson, “160 Gbit/s clock recovery using electroabsorption modulator-based phase-locked loop,” Electron. Lett. 36, 1951 (2000). [CrossRef]
3. D. Chiaroni, A. Jourdan, T. Zami, J-Y. Emery, and L. Tancevsky. “Toward 10 Tbit/s optical packet routers for backbone,” in Proceedings of ECOC, (European Conference on Optical Communications, Munich, Germany, 2000), paper Th 10.4.7
4. B. Sartorius, “All-optical clock recovery for 3R optical regeneration”, in Proceedings Optical Fiber Communication, (Optical Society of America, 2001), paper MG7-1.
5. T. Miyazaki, H. Sotobayashi, and W. Chujo, “Synchronous optical demultiplexing and sampling of 80 Gb/s OTDM signals by optically recovered clock using mode-locked laser diode and symetric Mach-Zehnder switch,” Photon. Technol. Lett. 14, 1734 (2003). [CrossRef]
6. M. Jinno and T. Matsumoto, “Optical tank circuits used for all optical timing recovery,” J. Quantum Electron. 28, 895 (1992). [CrossRef]
7. G. Contestabile, A. D’Errico, M. Presi, and E. Ciaramella, “40 GHz all optical clock extraction using a semiconductor assisted Fabry-Perot Filter,” Photon. Technol. Lett. 16, 2523 (2004). [CrossRef]
8. V. Roncin, B. Le-Guyader, S. Lobo, B. Clouet, and J-C Simon, “43 Gbit/s bit error rate assessment of a simple All Optical Clock Recovery scheme,” in Proceedings of ECOC, (European Conference on Optical Communications, Glasgow, Scotland, 2005), paper Th1.3.7, pp. 201–202
9. Ch. Kouloumentas, A. Tzanakaki, and I. Tomkos, “Clock recovery at 160 gb/s and beyond, using a fiber-based optical power limiter,” Photon. Technol. Lett. 18, 2365 (2006). [CrossRef]
10. I. D. Phillips, A. D. Ellis, T. Widdowson, D. Nesset, A. E. Kelly, and D. Trommer, “100 Gbit/s optical clock recovery using electrical phaselocked loopconsisting of commercially available components,” Electron. Lett. 36, 650 (2000). [CrossRef]
11. T. Yamamoto, L. K. Oxenlowe, C. Schmidt, C. Schubert, E. Hilliger, U. Feiste, J. Berger, R. Ludwig, and H. G. Weber, “Clock recovery from 160 Gbit/s data signals using phase-locked loopwith interferometric optical switch based on semiconductor opticalamplifier,” Electron. Lett. 37, 509 (2001). [CrossRef]
12. C. Bornholdt, J. Slovak, H. J. Thiele, and B. Sartorius, “Optical versus electrical clock recovery at 40 Gbit/s,” in Proceedings of ECOC, (European Conference on Optical Communications, Stockholm, Sweden, 2004), paper We4 P.080
13. C. Bintjas, K. Yiannopoulos, N. Pleros, G. Theophilopoulos, M. Kalyvas, H. Avramoloulos, and G. Guekos, “Clock recovery circuit for optical packets,” Photon. Technol. Lett. 14, 1363 (2002). [CrossRef]
14. V. Roncin, S. Lobo, L. Bramerie, and J-C. Simon, “Phase noise reduction in All Optical Clock Recovery at 43 Gb/s for 3R regeneration applications,” in Proceedings of ECOC, (European Conference on Optical Communications, Cannes, France, 2006), paper We3-P-91, pp.303–304
15. G. P. Agrawal and N. K. Dutta, “Semiconductor Lasers,” Second Edition, (Van Nostrand Reinhold, New York, 1993)
16. X. Zhou, C. Lu, P. Shum, H. H. M. Shalaby, T. H. Cheng, and P. Ye “A performance analysis of an all-optical clock extraction circuit based on Fabry-Perot filter,” J. Lightwave Technol. 19, 603 (2001). [CrossRef]